Back in February or March I applied for a senior staff role at LinkedIn, something pretty close to what I'm doing now. The first round was behavioral, and I've always felt I hadn't quite prepared it right — I leaned too hard on execution. I interviewed at Stripe recently too and got the same feedback. In the end they told me they wanted to downlevel me to staff. LinkedIn's senior staff, on levels.fyi, sits about half a level above staff at companies like Google or Meta — closer to Amazon's L7 — while staff itself sits about half a level below Google/Meta staff. I wasn't in a rush to leave, so I was ready to just let it go — LinkedIn's staff median comp on levels.fyi is noticeably lower than my current package anyway — but the recruiter kept telling me their staff comp could match what I'm making now, maybe even a bit more, and talked me into continuing on for staff. I agreed. Then she came back and said this particular team didn't have a staff opening in Seattle, so I got put on hold.
Then about a month ago the recruiter reached out again, saying there was now a staff opening in Seattle, but on a different team — this one working on CI/CD and developer experience. She asked if I wanted to interview. I figured since I'd already prepped, why not, even though it wasn't really a match for my background.
Round 1, behavioral. The manager interviewing me showed up about ten minutes late. Since I'd changed my behavioral strategy after the earlier feedback, I felt like this time I did a good job showing G/M-staff-level scope (comparable to LinkedIn's senior staff scope). He told me that wasn't right — that was senior staff scope, so why was I interviewing for staff? I explained the whole backstory. He said if he thought I was at senior staff scope, I could ask to be referred to senior staff, but senior staff on his team needs domain expertise I don't have, so it would be hard to bring me in as senior staff — though he also wasn't sure I'd enjoy the staff role either. I got the sense he was hinting I should just drop out, but I figured I'd already come this far, might as well get something out of it, so I said let's keep going. He said, well if you're going to keep going, I need to ask you some domain knowledge questions — and then the interview completely shifted from behavioral to questions about CI/CD and package management. I was pretty thrown off. Some of it I wasn't very familiar with and he didn't seem satisfied with my answers; some of it was closer to what I work on now and I answered well, and he'd say "good" — but out of everything he asked, I didn't get many "good"s. I'm guessing this round didn't go well.
Round 2, coding with AI. The problem was really convoluted, with a bunch of words I didn't even recognize — airdrop, retreat, stuff like that. But the first part, underneath the flavor text, was basically: there are a bunch of locations connected by paths, some locations have supplies, and when a paratrooper lands you need to tell them how far the nearest supply point is — essentially a graph traversal problem. There were two parts; I don't remember the second part very well. This interviewer confused me too — my usual approach for AI coding interviews is to let the AI come up with the algorithm and write the code while I review and evaluate it. I did the first part that way with no issues, but on the second part the interviewer suddenly told me not to use AI and to come up with the algorithm myself. Wasn't this supposed to be an AI coding round? The second part was harder than the first — you had to reconstruct the graph into a tree — and I thought about it for a long time without landing on a solid approach. I finally told him I'd just go ahead and use AI after all. I'm guessing the evaluation on this one wasn't good either.
Round 3, system design — build a job scheduler for a CI system. I'm still not that familiar with CI itself, but on the job-scheduler part alone I think I did fine. He also asked a few CI-specific things, like how to get CI build stdout to show up in the UI, and I didn't answer that well. There was also a whole back-and-forth at the start about whether GitHub pushes changes to the CI system or the CI system pulls changes from GitHub — he insisted pull was the right answer, but in all my years doing this I've never heard of a CI system pulling changes from GitHub. After the interview I actually looked it up, and sure enough, modern systems push: GitHub calls a CI endpoint to trigger the CI when there's a change. Pull is only used by old legacy systems, or private CIs that don't want to expose an endpoint. I really couldn't stand him. I'm guessing this round didn't go well either.
Round 4, regular coding — a robot-navigation problem. You're given some APIs to control a robot in an unexplored grid: turn, move forward, check whether there's a wall, and move the robot to any cell that's already been explored. You need to find a path from the robot's starting point to the destination. It's essentially BFS, except since the map hasn't been explored yet, you have to keep using the robot to probe the surroundings for walls and passability. I think this round went well.
It's been over a week now with no word, so I'm assuming I didn't get it. Honestly, though, I wasn't that excited about going in as staff on that team anyway — kind of like how I wasn't that excited about going in as an L6 at Amazon either. Getting downleveled by half a level makes the next round of level negotiations at my next job hop even harder.
Discussion
Loading comments…